提交 3c8aa787 编写于 作者: X xuezhong

define sampled_softmax_with_cross_entropy

上级 15d52f09
......@@ -87,7 +87,7 @@ __all__ = [
'transpose',
'im2sequence',
'nce',
'sample_logits',
'sampled_softmax_with_cross_entropy',
'hsigmoid',
'beam_search',
'row_conv',
......@@ -5765,23 +5765,22 @@ def softmax_with_cross_entropy(logits,
return loss
def sample_logits(logits,
label,
num_samples,
uniq=True,
remove_accidental_hits=True,
use_custom_samples=False,
custom_samples=None,
custom_probabilities=None,
seed=0):
def sampled_softmax_with_cross_entropy(logits,
label,
num_samples,
num_true=num_true,
remove_accidental_hits=True,
use_custom_samples=False,
custom_samples=None,
custom_probabilities=None,
seed=0):
"""
**Sampled Softmax With Cross Entropy Operator.**
Cross entropy loss with sampled softmax is used as the output layer for
larger output classes extensively. This operator samples a number of samples
for each example(row), and computes the softmax normalized values for each
for all examples, and computes the softmax normalized values for each
row of the sampled tensor, after which cross-entropy loss is computed.
This provides a more numerically stable gradient.
Because this operator performs a softmax on logits internally, it expects
unscaled logits. This operator should not be used with the output of
......@@ -5810,13 +5809,19 @@ def sample_logits(logits,
labels per example.
num_samples (int): The number for each example, num_samples should be
less than the number of class.
seed (int): The random seed for generating random number, which is used
in the process of sampling. Default is 0.
num_true(int): The number of target classes per training example.
remove_accidental_hits (bool): A flag indicating whether to remove
accidental hits when sampling. If True and if a sample[i, j]
accidentally hits true labels, then the corresponding
sampled_logits[i, j] is minus by 1e20 to make its softmax result
close to zero. Default is True.
use_custom_samples (bool): Whether to use custom samples and probabities to sample
logits.
custom_samples (Variable): User defined samples, which is a 1-D tensor with shape [S]. S is the num_samples.
custom_probabilities (Variable): User defined probabilities of samples, a 1-D tensor which has the same shape with custom_samples.
seed (int): The random seed for generating random number, which is used
in the process of sampling. Default is 0.
Returns:
Variable: Return the cross entropy loss which is a 2-D tensor with shape
......@@ -5855,12 +5860,21 @@ def sample_logits(logits,
},
attrs={
'use_custom_samples': use_custom_samples,
'uniq': uniq,
'uniq': True,
'remove_accidental_hits': remove_accidental_hits,
'num_samples': num_samples,
'seed': seed
})
return sampled_logits, sampled_label, samples, probabilities
helper.append_op(
type='softmax_with_cross_entropy',
inputs={
'Logits': sampled_logits,
'Label': sampled_label,
'soft_label': False,
},
outputs={'loss': samples, })
return outputs / num_true
def smooth_l1(x, y, inside_weight=None, outside_weight=None, sigma=None):
......
......@@ -374,7 +374,7 @@ class TestBook(unittest.TestCase):
self.assertIsNotNone(output)
print(str(program))
def test_sample_logits(self):
def test_sampled_softmax_with_cross_entropy(self):
program = Program()
with program_guard(program):
logits = layers.data(name='Logits', shape=[256], dtype='float64')
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册